| Age | Commit message (Collapse) | Author |
|---|
|
This was a nasty one. By closing the up channel, the up() function
would finish and send to the done channel. This means that the select
between err and done would be random as to which was picked, whereas
of course if there has been an error that path must be taken.
|
|
|
|
|
|
|
|
|
|
There are several ways that disk usage is reduced with this patch:
- Files are deleted as soon as they have been uploaded
- Once a page image has been added to a PDF, immediately delete it
This should allow much larger books to be processed without needing
bigger disks.
|
|
|
|
|
|
|
|
the cloudsettings documentation slightly
|
|
|
|
|
|
|
|
(which reenables it for spot instances)
|
|
HeartbeatSeconds, as it is not a Time
|
|
ensure it is always a / delimeter
|
|
|
|
|
|
|
|
minutes
|
|
|
|
|
|
|
|
|
|
|
|
permissions
|
|
This simplifies things nicely from using conn.GetLogger().Println() to
conn.Log()
|
|
Embarassing I hadn't noticed the fmt.Errorf function before, but
better late than never.
|
|
analysis as failed if graph isn't made for that reason
|
|
|
|
|
|
|
|
|
|
|
|
|
|
allOCRed was checking for wipePattern files, however they should have
been transformed into the regular preprocessedPattern for OCR anyway,
so shouldn't have been directly OCRed. Thus, allOCRed was mistakenly
looking for .hocr versions of the original wipePattern files, which
never would have been produced.
|
|
|
|
parsed by graph, and add line or word option
|
|
|
|
|
|
|
|
|
|
|
|
|
|
to any queue
|
|
output
|
|
|
|
|
|
|
|
|
